On March 8, we released significant revisions to the liability catastrophe model available to Oortfolio users. The modeling and data enhancements in this release include the following:
- Revisions to the state-transition model governing the flow of plaintiffs from exposure to injury to claim and settlement
- Relaxation of a purely deterministic allocation mechanism for distributing economy-wide losses to individual companies
- A revision to the method for allocating ground-up losses across policy years under the occurrence form that accounts for the most common allocation methods employed by U.S. courts today
- A revision to the decision rule governing the declaration of an integrated occurrence under the Bermuda form that accounts for the likelihood that an insured’s self-insured retention will be breached
- Incorporation of updated scientific (general causation) risk projections
- Incorporation of updated company profiles and newly profiled companies
Brief explanations of these modeling and data enhancements follow. Please contact your account manager for more detailed explanations.
Revisions to the state-transition model
Our liability catastrophe model simulates thousands of hypothetical mass litigation events. A key part of the model involves simulating the flow of plaintiffs from initial exposure to a Litagion® agent to disease manifestation to legal claim and settlement. Tracking the stocks and flows of individuals in each of these states over time is critical to estimating the time-path of settlements, defense costs, and indemnity payments under alternative insurance forms (e.g., claims made, occurrence, Bermuda Form). This model release includes a revision to how we estimate flows into and out of settings in which plaintiffs are exposed to a Litagion agent. The principal effect of this revision is to increase, by a small margin, the number of individuals who could ever assert they were exposed to a Litagion agent.
In addition, this model release now takes advantage of the full projected time-series of plaintiff case strength. Prior to this release, it was assumed that the final case strength value that governed claim and settlement behavior from year eight onward was equal to the average case strength between years one and seven of our projection. With this release, we now assume that case strength estimated in year seven of the projection will persist from that point forward. The effect of this revision is to increase the variability of case strength and, therefore, losses in the tail scenarios for most simulated litigation events.
This model release also smooths the surge in claims and settlements that occurs when the simulation first projects plaintiffs’ case becoming viable in court. In practice, mass litigation never results in a sudden flood of settlements in the first years of an event. Instead, mass litigation begins with a small number of bellwether cases that must make their way through the courts over a period of years before a mass settlement strategy emerges (if ever). This revision has the effect of spreading defense costs and indemnity payments over a larger number of years than was estimated in previous versions of the model.
Stochastic allocation of losses to modeled companies
This model release introduces variability in the share of economy-wide losses flowing to a modeled company. For a given event, economy-wide losses are first spread between distinct commercial activities according to the ease with which plaintiffs can attribute exposure to a company engaged in that activity and then by market share within the commercial activity. Prior to this release, these loss allocation shares were applied deterministically to arrive at estimated losses within any given event. With this release, these loss allocation shares are assumed to be random variables with a variance that is a function of the company’s estimated market share. This estimated variability generally increases losses in tail scenarios for any given company. The effect of introducing company-level variance on estimated tail losses, however, is typically minimal at the portfolio-level since each company’s loss share is independent of every other company’s loss share.
Revisions to estimated losses under the occurrence form
U.S. courts have employed a variety of methods over the years to determine which of many potential policy years must indemnify losses when the policy is issued on the occurrence form. These methods include, for example, the exposure, manifestation, continuous, and injury-in-fact trigger theories. Prior to this model release, our model of loss allocation under the occurrence form assumed one of two relatively extreme outcomes. Our upper-bound estimate of losses under the occurrence form [“occurrence (high)”] assumes that losses will be allocated between the current policy year and a future year in which insurers decide to exclude coverage for the risk whereas our lower-bound estimate [“occurrence (low)”] allows losses to be allocated over any policy year between first exposure and year of exclusion.
With this model release, we introduce a new estimate of losses under the occurrence form (labeled, simply, “occurrence”) that assumes a middle ground between occurrence (low) and occurrence (high). This estimate considers two key factors in determining loss allocation that vary widely between states: the selection of a trigger theory and the method of apportionment to implicated policies within the triggered policy periods. Loss estimates under the new occurrence form model are, on average, 150 percent greater than losses estimated under occurrence (low) and 60 percent less than losses estimated under occurrence (high).
The Oortfolio upload template has been modified to allow users to select the “occurrence” form option at the policy level when uploading portfolios.
Revisions to estimated losses under the Bermuda form
In order to estimate losses under the Bermuda form we must first determine whether there will be a declaration of an integrated occurrence (IO) in the policy year. If there is no IO declaration, then there will be no losses attributable to that policy year. Prior to this model release, an IO declaration was assumed to be a function of plaintiffs’ case strength alone, the idea being that the primary trigger for an IO declaration is the emergence of a litigation event. In our model, litigation can emerge even when plaintiffs’ case is still rather weak, and the magnitude of future mass settlements and whether an attachment point will be breached is uncertain. This modeling choice generated what could be considered to be an upper-bound estimate of IO declarations and hence losses under the Bermuda form.
With this model release, we now model the IO declaration as a function of the likelihood that a given insured’s self-insured retention (SIR) will be breached. The declaration is still critically a function of plaintiffs’ case strength, but it is also a function of the insured’s estimated losses given that case strength. Thus, the size of the economy-wide litigation event matters (which is a function of number of plaintiffs, their damages, etc.) as does the allocation of those economy-wide losses to the insured (a function of market share) and how those estimated losses compare to the insured’s SIR. The result of this more nuanced modeling approach is to reduce significantly the likelihood of an IO declaration in the current policy year.
Updated general causation risk score projections
This model release incorporates general causation risk scores and their projections as of October 2017. Estimated case strength is strongly influenced by general causation risk (the likelihood plaintiffs can convince a court that their injury could have been caused by exposure to the Litagion agent) and so these updated projections have resulted in significant revisions to both economy-wide and company-specific estimates of mass litigation losses attributable to certain Litagion agents. Economy-wide losses under tail scenarios for formaldehyde, for example, have fallen considerably whereas tail losses for dibutyl phthalate have increased markedly as a result of these updated projections.
It is important to note that the case strength projections have changed not because of a change in projection methodology but because the science investigating hypotheses of harm for some Litagion agents has shifted direction. For example, the general causation risk scores for several hypotheses of harm related to bisphenol A have leveled-off or even declined recently. Accounting for this shift in science, the case strength projections for claims alleging those harms result in a lower probability of plaintiffs being able to sustain such claims in the future.
Updated general causation risk score projections in combination with the model revisions noted above have resulted in the addition of 14 Litagion agents to Oortfolio whose economy-wide losses were previously inconsequential. These Litagion agents are now available for analysis within Oortfolio and are listed in the Oortfolio upload template.
Updated company profiles
This model release incorporates updates to analyst-profiled companies and the addition of 74 newly profiled companies. Algorithmically-profiled companies are unchanged. Newly profiled companies are now available for analysis within Oortfolio and are listed in the Oortfolio upload template.